Lab exercise 02, Supervised learning, KNN

1, Load and preprocess the data

• A, Download the modified version of the “Sleep in Mammals: Ecological and Constitutional Correlates” dataset. https://patbaa.web.elte.hu/Mammals.xls
• B, Check manually the excel sheet, which value marks the missing values?
• C, Load the excel sheet directly via pandas (the ‘All Data’ sheet). Tell pandas to track the missing values correctly!
• D, Remove columns that have more than 10 missing value!
• E, Replace each remained missing value with the mean of its column!

2, K-nearest neighbors regression

• A, Predict the lifespan of each animal with KNN regression. Use Euclidean distance, uniform weights in your model and set K to 10. Generate the predictions with leave-one-out cross-validation.
• B, Plot the predicted lifespan for each aminal vs the actual lifespan.
• C, Calculate the Pearson’s correlation and the MAE (mean absolute error) between the predicted and the actual values.

3, K-nearest neighbors regression with normalization.

• A, Normalize each column of your data frame (except for the predicted variable, the lifespan) to have 0 mean, and unit variance, and repeat exercise 2.
• B, Why is scaling of the data is important when using KNN? Write down with your own words!
• C, Investigate the predictions. Which animals die too early/live too long according to our model?
• D, Plot the MAE (mean absolute error) of the prediction as a function of K for K $\in$ {1; 5; 10; 15}

4, K-nearest neighbors regression by hand. You may use only numpy and pandas for this task.

• A, Calculate the lifespan KNN (10 neighbor, Euclidean distance, uniform weights) prediction for the Man. Use the normalized dataset. Fit the model on all the data except for the Man.
• B, How does your hand-crafted prediction matches with the prediction generated in 3,A?

Hints:

• Decorate your notebook with, questions, explanation etc, make it self contained and understandable!
• Comments you code
• Write functions for repetitive tasks!
• Use the pandas package for data loading and handling
• Use matplotlib for plotting or bokeh and plotly for interactive investigation
• Use the scikit learn package for almost everything
• Use for loops only if it is really necessary!